Tuesday, October 10, 2017

Ada Lovelace Day 2017

It is sad that my Ada Lovelace post this year has to be for Maryam  Mirzakhani. I never knew she was ill and had hoped to catch some impressive-named lectures from her in Stanford, sooner or later. I thought I had plenty of time. And Stanford has some pretty amazing lecture series asking for general audience kinds of talks. The woman was really impressive, it is a real shame to see her gone so young.

And yes this has been such a bad year for mathematicians, it feels spooky. Fred Linton  and Vladimir Voevodsky (I need to read the long thread in the Coq mailing list) in September and Miles Tierney in October.  But, personally for me, the hardest, Mike Gordon in  August. So very sad!

Monday, September 25, 2017

Meaning in Context?

So this was already the third meeting on Meaning in Context, but I finally got to hear about it.
Annie sent an invitation to Meaning in Context (MIC 3)
and I choose to join the working group on
Neural networks and Textual Inference, animated by Lauri Karttunen and Ignacio Cases.
What I did not understand was that I supposed to come to all the other talks too. Or maybe I wasn't. But the two days that I went  there were really very informative. Learned lots, as usual. 
The previous Meaning in Context were in Dagstuhl (edited by Hans Kamp, Alessandro Lenci and James Pustejovsky) and Munich (edited by Hinrich Schutze).
I finally met Ido Dagan and Jason Baldridge, which was nice. Now I need to see whether and if so,  how, can we move the work with  Livy and Katerina in such way as to use some of the ideas I hope we can borrow.

Friday, July 14, 2017

Logic at LiCS2017

Since there are no pictures of me talking at the Logic Mentoring Workshop (LMW) the mural at Keflavik Airport in Iceland will have to do.

I talked about "Weapons of Math Construction" (slides) and was pleased that after a certain amount of to-and-froing with the organizers, we were all happy with the event.

Also very happy with the quality of the talks in the "Women in Logic Workshop (WiL2017).

Not so happy with the discussion in the LiCS Business Meeting. As the SIGLOG chair Prakash Panangaden says in the SIGLOG news:

"At LICS in Reykjavik, Valeria de Paiva made a passionate plea for greater diversity in the LICS community. I think it is fair to say that not everyone was sympathetic. I don’t have the answers but certainly this issue cannot be ignored; it should be discussed widely.
I think it would be a great idea if people were to write to me or to the Editor so we can publish some of the opinions in the following two or three issues". 

I'm agreeing heartily with Prakash, everyone should write to him!

Thursday, July 13, 2017

Seven Different Valerias?

In some countries the legend is that cats have nine lives, in others  only seven.
Apparently Microsoft Academic (new version 2.0 just out!) thinks I am a cat.
At least there are seven Valerias de Paiva, when I search and they are all ME.

Check it out!
One of them has a paper with 144 other authors on Universal Dependencies, really cool.

Thursday, June 15, 2017

Missing My Mentors

Oscar Wilde famously said: "To lose one parent may be regarded as a misfortune; to lose both looks like carelessness."

I wonder if the same applies to career mentors.

On the 5th June, I went to two memorial events, one at Stanford for Solomon Feferman, who died last year and another  one for Danny Bobrow who was my mentor and collaborator at Xerox PARC for almost nine years. Danny died on 20th March 2017. It hurts. It does get you down.

Also missing Grisha Mints!

Tuesday, June 6, 2017

Sisterhood in Logic....

The First Workshop for Women in Logic is about to happen in Reykjavik, Iceland.

The website is  Women in Logic (WiL) Workshop19th June 2017.

The program is looking great!

(this is the new banner that my special advisor Zach Sanchez  just produced, thanks Zach!)

the previous not optimal picture was

Friday, May 12, 2017

Universal Dependencies for Textual Inference

In February 2017 I gave a talk at Nuance's AI Lab about a small experiment that I did with Alexandre Rademaker and Fabricio Chalub, from IBM Research, in Rio de Janeiro.

We used the SICK corpus created by the COMPOSES project in Italy, (devised to downplay the difficulties of language understanding), together with Parsey McParseFace,  Google's self-declared most accurate model in the world, to create logical representations of the sentences in SICK.

I reasoned that with all the fanfare about the advances of neural nets in NLP, considering that the corpus is simplified on purpose, these representations should be accurate enough to allow us to do the inferences required. Unfortunately between small errors here and there and big errors in the disambiguation, this experiment did not work the  way I expected it to.

You can see the slides in slideshare. Since the disambiguation, using Freeling's version of personalized PageRank, didn't work at all, we have all the possible word senses from WordNet at the moment in the GitHub repository. Now I am thinking about disambiguation, but also thinking about the kinds of inferences that we want and don't want to make. This project was suggested by Danny Bobrow several years ago.